Font Size: a A A

SOC Estimating Of Ni/MH Battery Band Using Neural Network

Posted on:2008-02-11Degree:MasterType:Thesis
Country:ChinaCandidate:S YouFull Text:PDF
GTID:2132360245992842Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Many countries around the world now spend a lot on the research of electric vehicle with the hope that it could completely replace the fuel vehicle one day for the pressure that environment pollution and energy crisis imposed on them. However, many critical technologies are still demanded to be solved to develop electric vehicle, among which is the battery management system. While the strategy of estimating the state of charge (SOC) of batteries is the most important technology in the battery management system, which contributes to the main task of my research.In this paper, we first defined the equation to calculate SOC of a battery, and analyzed all the variables in the equation especially which lead to the difficulties of estimating SOC. Then, a comprehensive summarize on SOC estimating arithmetic was presented, which showed both the merits and the defects about them, and their application in the real world. It also pointed out the future of SOC estimation.In this design, we accomplished many constant current charge/discharge experiments and many pulsed current charge/discharge experiments against many current values, and got lots of sample data. Some of these samples named traning sample were used to train the neural network, and others named testing sample were used to test the network we have constructed. These experiments were all done on the battery testing platform which was developed by Tianjin univercity, while the experimental data were acquired using high-precision and high-sample frequency ADC of dSPACE.Lately, this paper presented a brand new arithmetic to estimating SOC which combines the neural network method and the Ah method. BP neural network uses the current value and present SOC to forecast next value of charge/discharge coefficiency, and then SOC is figured out by Ah method which just needs the correct value of charge/discharge efficiency. The training result showed that the BP neural network has good generalization ability, while the simulation of SOC calculatimg model showed a satisfying result.At last, this paper analyzed some defects of this design, and pointed out how to solve them in the further research.
Keywords/Search Tags:State Of Charge, Ni/MH Battery, Back-Propagation Neural Network, Electric Vehicle, Hardware In-The-Loop Simulation
PDF Full Text Request
Related items